Pluto and eSpeak for ARMv7, now open source
Pages: 1 2
Martin Avison (27) 1494 posts |
Impossible to say that it (or any other software) has none. I can say that v3.16 has 14 less Zero Page problems than v3.14, and in normal daily use I (and other beta testers) have not seen one for months. But every possible obscure path through the program has not been tested … so there could still be some, but certainly off the beaten track. |
Chris Evans (457) 1614 posts |
I think the problem is worse than that. AIUI the null pointer dereferencing problem often only occurs if a routine is called with a particular value (zero?) in the register being used? Or have I misunderstood things? ZPP is great for debugging but I think it should never go into a ‘release’ build. |
Martin Avison (27) 1494 posts |
Only a little – it is when any code causes memory in page zero (ie 0 to 4096) to be read or written, not just a routine call. But you are right that it impossible to test for (at least without massive resources). |
Jon Abbott (1421) 2651 posts |
It’s fine for a release build, provided software is bug fixed/rewritten to not touch Page Zero. The problem with relying on null dereferences etc to work, is that a program will appear to be working correctly but may well be going horribly wrong because the Page Zero read/write is unpredictable. I’ve seen this issue in many games that I’ve fixed Page Zero bugs in over the past year. |
Pages: 1 2